PrismML Compresses Qwen3.6 to 4GB, Fits 27B Model into iPhone

According to Dynamic Beating monitoring, PrismML, an AI lab co-founded by Caltech mathematician Babak Hassibi, has successfully run the Alibaba open-source model Qwen3.6-27B compressed on an iPhone 17 Pro.

Qwen3.6-27B is a 27-billion-parameter dense model where all parameters are involved in inference, unlike MoE models that only activate certain parameters.

PrismML stated that the model size has been compressed from around 54GB to less than 4GB with no significant loss in performance. It can be used for complex dialogues, inference, agents, and code-related tasks.

Apple has been in touch with PrismML to discuss the implementation of this technology. Apple has been looking to incorporate more AI locally on devices, but advanced Siri still relies on cloud-based models.

Source #Cutting-edge trends #Ali.#apple

Correction/Report

RelatedRecommended

Apple earnings breakdown: iPhone drives growth, while profits are weighed down by storage and AI

The services business is positioned higher in the gross margin statement than in the revenue statement, and the growth rate in Greater China is faster than in Apple's other major reporting regions.

The company that makes glass for the iPhone saw its stock price increase fivefold due to AI.

Over the past 6 months, the company's stock price has increased by 140%, and compared to two years ago, it has multiplied by 5 times.

How Can iPhone Price Increases Avoid Hurting Sales? JPMorgan Breaks Down Apple's 'Monthly Installment Strategy'

Pro price hikes boost profits, foldable Duo may trigger a wave of device upgrades.

A conversation with Coinbase founder Brian Armstrong: Crypto’s iPhone moment has arrived

All the infrastructure is in place, but the user experience hasn’t really tied it all together yet, and there’s a huge opportunity here.

Web3 wallet "iPhone Moment", understanding the OuiAA smart contract wallet in one article.

This article aims to help new users understand and try to use Web3 wallets in the most straightforward language possible, so that they can join the new wave as soon as possible.

On-Chain Activity